US8174978B2 - Method for congestion management of a network, a signalling protocol, a switch, an end station and a network - Google Patents
Method for congestion management of a network, a signalling protocol, a switch, an end station and a network Download PDFInfo
- Publication number
- US8174978B2 US8174978B2 US10/590,209 US59020905A US8174978B2 US 8174978 B2 US8174978 B2 US 8174978B2 US 59020905 A US59020905 A US 59020905A US 8174978 B2 US8174978 B2 US 8174978B2
- Authority
- US
- United States
- Prior art keywords
- port
- congestion
- ingress
- data packets
- egress
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active, expires
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000011664 signaling Effects 0.000 title description 8
- 238000011144 upstream manufacturing Methods 0.000 claims abstract description 118
- 239000004744 fabric Substances 0.000 claims description 23
- 230000005540 biological transmission Effects 0.000 claims description 10
- 230000004044 response Effects 0.000 claims description 8
- 230000015572 biosynthetic process Effects 0.000 claims description 3
- 230000008878 coupling Effects 0.000 claims description 2
- 238000010168 coupling process Methods 0.000 claims description 2
- 238000005859 coupling reaction Methods 0.000 claims description 2
- 239000003550 marker Substances 0.000 claims description 2
- 230000000694 effects Effects 0.000 description 22
- 239000000243 solution Substances 0.000 description 20
- 230000000903 blocking effect Effects 0.000 description 15
- 238000007726 management method Methods 0.000 description 11
- 230000003213 activating effect Effects 0.000 description 7
- 230000008901 benefit Effects 0.000 description 4
- 230000007480 spreading Effects 0.000 description 4
- 239000000872 buffer Substances 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 235000008694 Humulus lupulus Nutrition 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000001934 delay Effects 0.000 description 2
- 238000002347 injection Methods 0.000 description 2
- 239000007924 injection Substances 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000001902 propagating effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000135 prohibitive effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/17—Interaction among intermediate nodes, e.g. hop by hop
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/26—Flow control; Congestion control using explicit feedback to the source, e.g. choke packets
- H04L47/266—Stopping or restarting the source, e.g. X-on or X-off
Definitions
- the present invention relates to a method of congestion management within a switch or a network of connected switches.
- the invention also relates to a switch, an endstation, a network and a signaling protocol for managing congestion within a network of switches.
- the invention relates to a method of congestion management within a single switch or within a network of connected switches, including the corresponding endpoints. In embodiments the invention also relates to the switch architecture and a signaling protocol for managing congestion within a network of one or more switches.
- Contention within a network occurs when several data flows across a network contend for access to the same egress port of a switch within the network. If the bandwidth provided by the egress port is not high enough to service all requests made of it then the contention becomes congestion. Data packets that cannot make forward progress are buffered in queues. If congestion persists those queues are filled. If filled or blocked queues are not prevented from forming, or if action is not taken to alleviate the blockages as they form then this leads to congestion which can spread rapidly throughout the network between connected switches, forming what is known as a congestion tree. This can cause substantial inefficiencies in large networks and may even result in total fabric collapse.
- HOL blocking occurs whenever a queue stores data packets for different destinations and the packet at the head of that queue is prevented from being forwarded because its destination is unable to accept the packet. This is usually a consequence of several flows contending for access to the same egress port of a switch within the network. These aggregation effects apply both internally to a switch and between the switches in a fabric. Subsequent packets in a blocked queue, i.e. packets behind the head of the line may be intended for different destinations which are able to accept the packets but these packets are still prevented from being forwarded by the blocked packet at the head of the line. As a consequence, throughput may degrade dramatically.
- the first group relies on modification of how end-station equipment injects traffic into packet switches or networks of switches. Solutions in this group usually modify the rate at which packets are injected into the network, and are referred to either as Network Admission Control, Congestion Control, or Injection Limitation techniques.
- the second group of solutions includes the techniques that directly avoid HOL blocking by providing separate buffers for different flows or extra hardware paths to avoid waiting for packets at the head of the line.
- VOQ Virtual Output Queuing
- Another known solution alleviates HOL blocking over only one hop or stage in a Multi-stage Interconnection Network (MIN). This is achieved by the establishment of a number of queues in an upstream switch which are usually pre-allocated in some fixed way to store packets for the egress ports of the next downstream switch.
- This solution requires signaling means between the downstream and upstream switches that is employed to provide status based flow control.
- This single stage limitation again leads to HOL blocking effects arising upstream from the switch that is responsive to the flow control information and to congestion spreading across the switching stages in a fabric.
- the method comprises at said upstream port creating an entry in a control memory, e.g. an associative memory or CAM, to indicate that congestion has occurred at the congested port; and,
- a control memory e.g. an associative memory or CAM
- the method comprises at said upstream port, allocating memory for use as the set-aside-queue for data packets destined for the congested port.
- the method comprises de-allocating one or more set aside queues in dependence on one or more criteria such as the amount of data in the set aside queue or whether a token has been received by a port, CAM line or switch. This will be explained below.
- the request when a request for storage of data packets received at any of the ports in the congestion tree is in respect of congestion at a port further downstream than the root of the congestion tree, the request is accepted such that data packets destined for said further downstream port are stored at the port at which the request was received
- a signaling protocol for managing congestion within a network of switches comprising:
- a switch for use in a network of switches comprising:
- a switch fabric for selectively coupling data packets received at one or more of the ingress ports to one or more of the egress ports;
- selection means for selectively routing a received data packet to the storage in dependence on the detected desired destination of the packet
- request generation means arranged to send a request to a further upstream port to request storage of data packets destined for the congested port at said further upstream port when a threshold amount of data packets are stored in the storage.
- a network of interconnected switches comprising a plurality of switches arranged in a topology, the network comprising at least two switches according to the third aspect of the present invention, the at least two switches being connected directly to each other.
- a switch for use in a network of switches comprising:
- control means for selectively routing data packets received at one or more of the ingress ports to one or more of the egress ports;
- At least one of the ingress ports or egress ports comprises storage for storing details of a congestion tree comprising at least three connected ports in which in use, the switch is located.
- the invention provides a solution to the Congestion Management problem.
- the invention comprises a set of means and a method to manage network admissions and reduce or eliminate congestion spreading due to HOL blocking effects over several hops of a MIN with any topology. This is achieved by preventing HOL blocks from forming.
- the invention enables network end stations and single stage switches to make more effective use of their queuing resources and also enables a multi-stage switch fabric to behave substantially as if it were a large single stage switch.
- a signaling protocol for managing congestion within a network of switches comprising:
- a second message for sending by the upstream port to a port further upstream when a threshold amount of data packets destined for the congested port have been received and stored by said upstream port, said message requesting storage of data packets destined for the congested port received by said further upstream port.
- said upstream port and said further upstream port are controlled to allocate a set aside queue at said upstream port or at said further upstream port respectively for storage of data packets destined for the congested port.
- an endstation for use in a network of interconnected switches, the end station comprising:
- an ingress port for receiving data packets from a network to which in use the end station is connected;
- an egress port for providing data packets to a network to which in use the end station is connected;
- the egress port comprises means operable in use to receive a message from a downstream port, the message containing data relating to a congested port downstream and a request to provide storage for data packets destined for the congested port downstream.
- the invention provides a scaleable solution to the Congestion Management problem.
- the invention comprises a set of means and a method to manage network admissions and reduce or eliminate congestion spreading due to HOL blocking effects over several hops of a multi-stage interconnect network (MIN) with any topology. This is achieved by preventing HOL blocks from forming.
- MIN multi-stage interconnect network
- the invention enables network end stations and single stage switches to make more effective use of their queuing resources and also enables a multi-stage switch fabric to behave substantially as if it were a large single stage switch.
- the invention uses a means to set aside traffic causing HOL blocking to allow traffic, which would otherwise be prevented from flowing through the switch or network by the blocked traffic, to flow on through the switch to unblocked network routings.
- the means includes a low bandwidth signaling method across the switch network to set aside traffic at the Head of Line in a queue, the method not requiring significant switch network bandwidth.
- the method is capable of responding quickly enough to prevent the entire switch network becoming locked up through the rapid escalation of congested traffic.
- the invention uses a fixed amount of control memory at each switch ingress/egress port regardless of the size of the switch fabric.
- the invention is applicable to network end stations, single stage switch elements and Multi-stage Interconnection Networks of switches, including regular and irregular interconnection patterns between stages and for fabrics of any size and topology.
- the invention provides a scalable and cost-effective solution to the Congestion Management problem.
- it solves the HOL blocking problem in a scalable manner by using a fixed amount of extra buffer resources in a particular way.
- the invention also provides benefits of the solutions in the group of Network Admission Control while avoiding network utilisation variation effects.
- This invention allows end stations to be properly responsive to a user specified network admission policy in a single stage or Multi-stage Interconnection Network (MIN). In operation it provides all the necessary information to enable end-stations to manage the rate at which they inject traffic for all destinations in a properly co-ordinated and stable manner.
- MIN Multi-stage Interconnection Network
- OEO out of order
- a sequence of data packets destined for the same congested port arrives at an upstream port, and in the time between arrival of the first and second of the data packets destined for the congested port the congestion clears it could be that the second data packet is sent to the now uncongested port before the first data packet.
- a preferred requirement is that, if a request for establishment of a set aside queue is received at a port and the request is in respect of a port further downstream than the furthest downstream already congested port, then this request is ignored.
- FIG. 1 shows an example of a network of interconnected switches the network comprising 8 input devices D 0 to D 7 connected to a three-stage network of switches (stages A, B and C).
- the network is configured such that each of devices D 0 to D 7 may send data to either or both of devices DA or DB.
- it is necessary to save the state of all the ingress devices D 0 to D 7 simultaneously.
- it may well be that all the state data at some point in time from the ingress devices D 0 to D 7 will be sent to device DA.
- congestion due to contention
- a 00 , A 10 , A 20 and A 30 congestion management method
- each ingress device D 0 to D 7 may, for example, only send data packets destined for DA at 25% of link capacity. The situation will be repeated at port C 0 , thereby limiting the effective transmitting capacity to DA to about 12% of link capacity.
- a stage switch egress ports 0 are seen as a congestion root, then data packets simply passing through the port A 20 will be assigned to a set aside queue for the ports of switch B 1 in switch A 2 , even if the data packets are destined for port B 11 which is not congested instead of port B 10 .
- the amount of link capacity used between other uncongested ports may unnecessarily be reduced.
- the step of requesting storage at the upstream port of data packets destined for the congested port comprises requesting establishment of a set aside queue for storage of said data packets; and data packets stored at said further upstream port are stored in a set aside queue for data packets destined for the congested port thereby establishing an original congestion tree; and when a subsequent request for storage of data packets is received at any of the ports in the original congestion tree in respect of congestion at a port further downstream than the root of the original congestion tree, the request is accepted at the port such that data packets destined for said further downstream port are stored at the port at which the request was received thereby extending the congestion tree downstream.
- the invention provides a means for enabling movement of the root of a congestion tree downstream. Accordingly, in networks of the type shown in FIG. 1 , network utilisation may be maximised.
- the invention provides a congestion management method for applications in which it is necessary to be able to move the root of a congestion tree downstream or rather to allow a congestion tree that develops further downstream than an existing root to exist simultaneously with the existing tree.
- 000 effects may arise if requests for establishment of set aside queues further downstream than an existing congestion tree root are accepted.
- the method comprises establishing one or more links between the set aside queue of data packets destined for the further downstream port and the set aside queue of data packets destined for one or more of the other congested ports in the congestion tree.
- the method preferably comprises accepting all requests for establishment of set aside queues, and when said requests are for establishment of a set aside queue in respect of a port further downstream than the root of the congestion tree, placing a link in one or more of the existing set aside queues to later activate the newly formed set aside queue.
- the method comprises: if a request is for establishment of a set aside queue in respect of a port further upstream than the root of the original congestion tree, overwriting the existing set aside queue having fewest stages with a newly established set aside queue; and
- the method comprising overwriting the existing shortest set aside queue with a newly established set aside queue corresponding to the received request; and placing a link to the newly established set aside queue in the already existing set aside queue that is the longest already existing set aside queue and that is shorter than the newly established set aside queue.
- FIG. 1 shows an example of a network of switches
- FIG. 2 shows a schematic representation of a conventional network of switches
- FIG. 3 shows a further example of a conventional network of switches
- FIG. 4A shows a schematic representation of an example of a switch according to an embodiment of the present invention
- FIG. 4B shows a schematic representation of an ingress port and an egress port within an example of a switch according to an embodiment of the present invention
- FIG. 4C shows a schematic representation of an example of an end station according to of an embodiment of the present invention.
- FIG. 5 shows a schematic representation of an example of a network of switches according to an embodiment of the present invention
- FIG. 6 shows an example of queues formed in a switch
- FIGS. 7 and 8 show schematic representations of an example of a network in accordance with an embodiment of the present invention.
- FIG. 9 is a representation of a content addressable memory and a number of set aside queues
- FIG. 10 is a representation of data queues at a port of a switch in a network of switches.
- FIG. 11 is a representation of data queues at a port of a switch in a network of switches.
- FIG. 2 shows a schematic representation of a network of switches 2 and 4 .
- switches 2 and 4 have a number of ingress ports A to C and a number of egress ports 0 to 2 .
- a data packet is received at an ingress port A to C of a switch and routed to one of the output ports 0 to 2 of the respective switch 2 and 4 .
- the network operates a local explicit congestion notification LECN protocol such that, for example, if a port on switch 4 becomes congested it sends a signal upstream to the switch port which is sending it data to temporarily stop transmission or to modulate the transmission in some way so that the blockage is able to clear.
- a signal is sent from port 0 of switch 4 to port 0 of switch 2 to instruct it temporarily to stop transmission.
- Data destined for port 0 of switch 4 is queued at port 0 of switch 2 in a set aside queue (SAQ).
- SAQ set aside queue
- FIG. 3 shows an example of a conventional network of switches 6 , 8 and 10 also utilising a LECN protocol as described above with reference to FIG. 2 .
- a notification is sent to port 0 of switch 8 telling that port to stop sending data to switch 10 that is destined for port 0 of switch 10 .
- port 0 of switch 8 subsequently becomes congested, it sends a similar notification to, for example, port 0 of switch 6 which blocks data from all input ports A to C of switch 6 intended for port 0 of switch 8 .
- data unrelated to the congestion for example, from port B of switch 6 to port 1 of switch 10 via port 0 of each of switches 6 and 8 is also blocked. Accordingly, head of line blocking occurs and a congestion tree develops. As explained above the creation of a congestion tree can have potentially serious consequences for operation of the entire network.
- FIG. 4A shows an example of a particular type of switch according to an embodiment of the present invention.
- the switch comprises a plurality of ingress ports 3 and a plurality of egress ports 5 .
- the ingress ports 3 are arranged to receive data packets from other switches within the network.
- the egress ports 5 are arranged to receive data packets from one or more of the ingress ports 3 and provide a route onward for the data packets.
- a switch architecture 7 is provided and is shown schematically and may be any suitable type of architecture controllable to couple data packets from one or more of the ingress ports 3 to a selected one or more of the egress ports 5 .
- An ingress engine 9 is provided in each of the ingress ports 3 .
- An egress engine 11 is provided in each of the egress ports 5 .
- the ingress engine 9 is operable to detect incoming data packets to the ingress port 3 , route the data packets to a particular egress port via an uncongested virtual output queue i.e. a cold queue 13 within the ingress port 3 or in some situations selectively route the received data packet via a SAQ 15 within the ingress port 3 .
- the egress engine 11 is operable to detect incoming data packets from an ingress port and selectively route them to a cold queue 17 within the egress port 5 or to a SAQ 19 within the egress port 5 .
- the operation of the ingress engine 9 and egress engine 11 will be described in more detail below.
- links between switches are duplex links, i.e. data can be passed directly in both directions between two directly connected switches.
- FIG. 4A data packets and control data are passed between ingress and egress ports within the switch by the switch fabric (core) shown schematically as a cross between the ingress and egress ports.
- any suitable means may be provided as the ingress engine 9 or the egress engine 11 for, amongst other functions, determining the routing of received data packets.
- the ingress engine 9 and egress engine 11 are provided by a content addressable memory (CAM). This has the advantage that it is implementable by hardware and therefore able to operate at a high rate. Typically, the mechanism must cope with link or port bit rates of up to hundreds of gigabits. In a network of switches overall throughput of many terabits could be achieved.
- the egress port 5 receives a request via its corresponding link ingress port from a downstream egress port of a switch within the network.
- the request is amongst other things a request to establish a SAQ 19 within the egress port 5 .
- the request is considered and checked against a number of requirements and if these requirements are met, the egress engine establishes SAQ 19 .
- the egress engine is then operable to determine the destination of an incoming data packet received from one or more of the ingress ports and selectively route the data packet either to the cold queue 17 within the egress port 5 or to the SAQ 19 .
- the ingress port 3 contains equivalent features and operates in a similar manner to the manner in which the egress port 5 operates. However, a request for establishment of an SAQ is only sent to an ingress port by an egress port when a threshold is reached in the egress port SAQ. Thus SAQ usage and notification traffic is minimised.
- the description above in relation to the ingress and egress ports 3 and 5 is of course a simplified description and a more detailed description will now be given of the protocol by which the ingress and egress ports operate.
- FIG. 4B shows a schematic representation of an ingress port and an egress port within one particular type of switch 50 according to an example of an embodiment of the present invention.
- the switch 50 has N+1 ingress ports and N+1 egress ports.
- ingress port 0 and egress port 0 are shown in any detail.
- an egress engine 11 is provided as described above with reference to FIG. 4A .
- the egress engine 11 is adapted to communicate with content addressable memory 62 and is operable to generate SAQs 64 in response to requests received via an ingress port from a port of a connected downstream switch.
- the engine 11 is also arranged for communication with cold queues 56 .
- the CAM 62 is arranged to communicate with SAQs 64 and also storage 66 provided for leaf tokens the purpose of which is described in detail below.
- SAQs 64 and also storage 66 provided for leaf tokens the purpose of which is described in detail below.
- egress port 0 sends a request to an upstream port requesting establishment of an SAQ at the upstream port for storage of data packets destined for a downstream port
- a leaf token is sent with the request and this is recorded in the storage or token memory 66 .
- the token is stored as a flag in an ingress port of the upstream switch. Owning a token allows a SAQ to be collapsed when certain conditions are satisfied, as described in further detail below.
- FIG. 4C shows a schematic representation of an end station according to an example of an embodiment of the present invention.
- the end station has a single ingress port 70 and a single egress port 72 containing an egress engine 74 .
- the end station operates in a similar manner to a switch described in detail herein except there is no selective routing of data packets between ingress and egress ports.
- the request is passed to the egress engine 74 within the egress port 72 .
- the egress engine functions in a similar manner to the egress engine described above with reference to FIGS. 4A and 4B . In other words, it functions to establish SAQs and appropriate entries in a CAM to enable selective routing of data packets to either a cold queue or a SAQ within the end station, in dependence on whether or not the data packets are destined for the congested port downstream from which the request originated.
- FIG. 5 shows a schematic representation of an example of a network of switches according to an embodiment of the present invention.
- a regional explicit congestion notification (RECN) protocol is used in addition to the LECN protocol used in and described with reference to the networks of FIGS. 2 and 3 .
- the network comprises three switches 76 , 78 and 80 .
- Each switch has three ingress ports A to C and three egress ports 0 to 2 .
- switch 76 At an end station or switch e.g. switch 76 , that is injecting traffic into a switch or network of switches and in an individual switch element a pool of dynamically allocated, associatively mapped SAQs are provided, as described above with reference to FIGS. 4A to 4C .
- these queues are operable in a manner responsive to a Regional Explicit Congestion Notification (RECN) protocol.
- RESN Regional Explicit Congestion Notification
- the RECN signaling protocol operates between and through the switches out to the end stations in a MIN to manage the allocation, population and de-allocation of the SAQs by flows which are persistently congested.
- the flows which are subject to this type of congestion are known as HOT flows and all other types of flow are referred to as COLD flows.
- Cold flows are always mapped to cold queues formed within the ingress or egress ports as described above with reference to FIGS. 4A to 4C , regardless of their destinations. Thus, buffer requirements are minimised.
- This strategy does not introduce significant HOL blocking because Cold flows are not blocked. When congestion is detected, and this may happen, for example, when a certain threshold is reached in a cold queue of a particular port, a notification is sent upstream that contains information about the congested port.
- the information on the congested port is compared by the CAM against previously stored notifications. In one embodiment, it is accepted only if it is unrelated or is more generic than previously stored notifications. Otherwise, the notification is discarded.
- a notification is accepted, a line in a memory such as a content addressable memory describing it is allocated. In addition, a corresponding SAQ is allocated.
- Incoming packets to the upstream switch are analysed and their intended paths compared against CAM lines.
- the packet In the case of a match, the packet is known to belong to a hot flow and will be stored in a corresponding SAQ either in an ingress port or an egress port depending on where SAQs have been established. Thus any potential HOL blocking that this packet could introduce is removed.
- flow control may be based on the transmission of Xon and Xoff, i.e. turn on and off, messages to upstream nodes.
- notifications propagate a token upstream.
- the token identifies a Leaf Node in a congestion tree i.e. a port of a switch within the congestion tree. All the leaves in the congestion tree will contain a token. Also, a record is kept at every egress port (from which requests are sent) to keep track of the number of tokens it has sent to upstream ingress ports within the current switch.
- a given ingress port When a given ingress port receives a de-allocation notification, it becomes the owner of the leaf token, and therefore becomes a leaf node. A given egress port must have all upstream tokens returned before it can become a leaf node. Only then can the corresponding CAM line become eligible to itself initiate the de-allocation mechanism, propagating the de-allocation notification to downstream switch ports, unless that particular switch port is the root of the congestion tree.
- CAMs are used to monitor the operation of the RECN protocol. It will be appreciated that CAMs are implementable with the use of hardware and therefore enable extremely fast control of steps in the RECN protocol. Of course, other means may be used to provide such control. In fact, any means may be used that is capable of identifying from an incoming packet whether it is destined for a congested port or not and routing it accordingly either to an SAQ or a cold queue.
- FIG. 6 shows a schematic representation of a cold queue and a SAQ as provided at a port of a switch in a network.
- the cold queue contains a list of data packets identified by the ports to which they are being sent.
- the CAM contains a number of lines 0 to 2 each containing an indication of packets that are in the SAQ CAM line details (not shown).
- a number of markers L B and L C are provided in the cold queue. These serve as links within the cold queue to maintain the chronological order of data packets. The markers serve to make visible the SAQ to an associated scheduler.
- the CAM compares their destination to information it is storing about congestion downstream. If it is determined that the packet is destined for a port known to be congested, the packet is directed to the corresponding SAQ and a marker is provided in the cold queue of the corresponding port. If however it is determined that the packet is not destined to a known area of congestion, the packet is not directed to the SAQ but rather routed directly to the cold queue of the port.
- direct network addressing methods at its origin, a packet is informed of its eventual destination. As the packet is received by switches en route to its destination, a look up table in the switch is referenced and from this look up table the appropriate egress port by which the packet should leave to continue its route to the destination is determined.
- deterministic source routing at its outset a binary instruction is given to a packet, the instruction containing an entire route for the packet from its origin to its destination.
- deterministic source routing or the use of turnpools is preferred due to the increased speed at which it is possible to perform routing operations.
- a turnpool consists of a set of contiguous variable sized fields in a packet header, and a pointer.
- the pointer is initialised by the source endpoint, and addresses the first valid field in the turnpool, which specifies the address of an egress port in the first downstream switch, relative to the ingress port.
- the size of the field is sufficient to enable the above addressing operation for all ports in the switch.
- a mask size mechanism referred to in CAM addressing operations provide a means of limiting the size (or range) of the turnpool fields used to address the CAM, and/or a means of identifying the size (or range) of the fields in the stored data.
- a fabric can consist of a series of connected switches that are smaller than the maximum size allowed by the turnpool size (which is limited to 31 bits). Alternatively, congestion can appear anywhere within the network. Therefore only the fields within the turnpool required to address the switches between the root and leaf nodes are pertinent to the CAM operations, and these must be determined at each CAM.
- An alignment operation is necessary as some paths through an irregular network are longer/shorter than others, i.e. take more/less turnpool bits to specify, so a known current position within the turnpool must be used for the alignment comparisons.
- a CAM is an example of a means that can be used to provide fast processing of data packets, i.e. positioning in SAQ or cold queue, in dependence on their destination.
- the CAM is able to process SAQ formation requests received from a downstream port and data packet assignments for data packets received from an upstream port.
- the CAM is arranged to determine if a received request is more specific than any existing entries in the CAM. This is achieved by detecting if any existing entries partially match the new request. If they do, the new request is more specific than the existing entry in the CAM. In one embodiment, such requests are rejected since this would refer to a port further downstream than the root of the congestion tree in which the port is located. As will be explained below this can lead to out of order effects.
- a further feature of the CAM is the line alignment for turnpool bits.
- active turnpool bits for all downstream switches to the congested port must be left aligned in the CAM. This enables correct determination of a packet's destination.
- RECN protocol is made up of four messages. These are:
- the port operating in accordance with the RECN protocol sends a request message to an upstream port (could be an ingress port or an egress port depending on where the congestion occurred) that is sending data packets to the congested port to request establishment of an SAQ.
- an upstream port (could be an ingress port or an egress port depending on where the congestion occurred) that is sending data packets to the congested port to request establishment of an SAQ.
- the upstream port is controlled to send an acknowledgement message to the downstream port from which the request originated, informing the downstream port that an SAQ has been established.
- flow control messages are sent by the downstream port to the upstream port in which the SAQ has been established, to control the flow of data packets from the SAQ to the port at which (or one stage closer to the point at which) the congestion has occurred.
- the flow control message may contain any one of a number of instructions such as, for example, stop transmission, start transmission, modulate transmission, etc.
- the final message in the RECN protocol is a de-allocation message sent by the upstream port to the downstream port informing the downstream port that the SAQ is being de-allocated. This might be for a number of possible reasons. Examples include that the SAQ has become empty and that a predetermined period of time has passed without any other data packet being received into the SAQ etc.
- a token is sent with the request message.
- the token identifies a leaf port in the congestion tree as mentioned above. Accordingly, all the leaves in the congestion tree will contain a token.
- a record is kept at every egress port within switches in the network to keep track of the number of tokens it has sent to upstream ports through different branches of the congestion tree.
- a given egress port When a given egress port receives a de-allocation notification, it waits until it has received all the tokens it previously sent to upstream ports. Only then, can the corresponding SAQ and CAM lines become eligible to initiate a de-allocation message, propagating de-allocation notifications to downstream switches, unless the switch itself is the root of the congestion tree. This provides a means for a port to know if it is eligible to de-allocate an SAQ or not.
- FIG. 7 shows a schematic representation of an example of a network of switches according to an embodiment of the present invention.
- SAQs are located at switch egress port 0 .
- the figure also shows CAM entries which refer to downstream switch ports or end points of the network.
- CAM lines 0 show contents for congestion source C 0 .
- CAM lines 1 show contents for congestion source E 1 .
- FIG. 8 shows a similar arrangement although in this case each of the switches only has two ingress and two egress ports.
- a LECN protocol is used to establish a SAQ at the egress port 0 of switch C.
- SAQs are established using the RECN protocol described above.
- CAM lines 0 relate to source of congestion C 0
- CAM lines 1 relate to source of congestion E 1 . It can be seen that as a switch gets further away from the source of congestion the entries in the CAM lines get more specific so that they can be used to identify packets the furthest upstream port, within the congestion tree all the way to the root of the congestion tree.
- an entry is created in a memory to indicate that congestion has occurred at the particular (further downstream) port.
- data packets are received at the port, they are checked against the entry or entries in the memory. If a data packet is directed to the congested port, the data packet is stored in the corresponding set aside queue.
- FIG. 9 shows a representation 82 of a CAM and the SAQs 84 , 86 , 88 and 90 formed based on the entries in the CAM. It can be seen that there are four entries in the CAM arranged respectively in rows 0 to 3 of the CAM.
- the SAQ 90 formed based on the A 1 B 1 C 1 in the CAM may be referred to as the “longest” of the SAQs in this example. It will be appreciated that the term “longest” does not refer to the number of data packets stored in the SAQ, but refers to the fact that the CAM entry that gave rise to it, is the longest, i.e. has the highest number of stages. Similarly, SAQ 88 may be referred to as the “shortest” SAQ in this example as it has the lowest number of stages (only a single stage).
- One SAQ 84 is formed to store data packets destined for the port B 1 along the route A 1 /B 1 .
- Another SAQ 86 is formed to store data packets destined for the port B 2 along the route A 1 /B 2 .
- Another SAQ 88 is formed to store data packets destined for the port A 1 , but not destined for either of the routes A 1 /B 1 or A 1 /B 2 .
- Last, SAQ 90 stores data packets destined for the port C 1 via ports A 1 and B 1 .
- FIG. 10 shows a representation of the sequence of arrival of data packets at a port in a network and the SAQs formed at the port.
- the figure shows the data packets arranged in a cold queue and a number of SAQs. Referring to the figure, data packets 0 to 44 are shown as arriving at the port. As each data packet arrives, its desired destination is checked against entries in the CAM. If there is a corresponding entry in the CAM the data packet is placed in the corresponding SAQ. If there is no corresponding entry in the CAM, the data packet is placed in the cold queue.
- data packets 5 to 9 , 15 to 19 , 25 to 29 etc. have not been shown. It can be seen that initially data packet 0 A 1 B 1 C 1 , arrives at the port, i.e. a data packet destined for port C 1 via ports A 1 and B 1 . This is followed by data packet A 1 B 1 C 2 . After packet 4 has arrived, a request A 1 B 1 arrives at the port. This is a request for establishment of a SAQ for data packets destined for Port B 1 via port A 1 .
- All data packets that have route A 1 B 1 as all or part of their designated routes will, at this stage in the absence of any other SAQs, be stored in the SAQ A 1 B 1 .
- data packets with routes A 1 B 1 C 1 , A 1 B 1 C 2 and A 1 B 1 C 3 etc will be stored in the SAQ A 1 B 1 .
- a SAQ A 1 B 1 is established and an activating link (R/L A 1 B 1 ) is placed in the cold queue at the port, linking the cold queue to the SAQ A 1 B 1 established in response to the request.
- Packets 10 and 11 arrive and these are placed in the established SAQ A 1 B 1 .
- Packets 12 to 14 are stored in the cold queue since there is no SAQ at present for any of their intended routes A 1 B 2 C 1 , A 1 B 2 C 2 and A 1 B 3 C 1 .
- a request for establishment for a SAQ for all data packets routed to or via port A 1 is received. This request is in respect of a port closer (i.e. fewer separating stages) to the port receiving the request than a port in respect of which there is an already existing SAQ. It is a “less specific” request.
- a CAM line A 1 is formed, as is a corresponding SAQ.
- a link to the SAQ A 1 is placed in the cold queue.
- Packets 22 to 24 that arrive subsequently are placed in the SAQ A 1 , since each of the packets is set to be routed via port A 1 , but not subsequently port B 1 .
- a request A 1 B 1 C 1 arrives for establishment of a SAQ for data packets destined for the route A 1 B 1 C 1 .
- This is a request in respect of congestion further downstream than the already existing farthest downstream congested port, i.e. further downstream than the root of the existing or original congestion tree.
- the request may be described as “more specific” than any of the existing SAQs (A 1 B 1 and A 1 ). Accordingly, a corresponding new SAQ A 1 B 1 C 1 is established and an activating link for the new SAQ A 1 B 1 C 1 is placed in an existing SAQ to avoid potential out of order effects.
- the link is placed in the longest existing SAQ, i.e.
- Packet 30 that subsequently arrives is now placed in the SAQ A 1 B 1 C 1 .
- SAQ A 1 B 1 C 1 there is no risk of out of order effects since it will only be transmitted after the packets 10 and 20 already in SAQ A 1 B 1 . It will be appreciated that by accepting more specific SAQ requests the root of the congestion tree is moved downstream.
- CAM lines are overwritten when a new request is received, and the new request either includes all stages of the route defined by the existing CAM line or is less specific than an existing CAM line.
- FIG. 11 shows a cold queue and a number of SAQs formed at a port in a switch in a network of connected switches. As in the example shown in FIG. 10 , for clarity and conciseness a number of data packets have not been shown.
- data packets 0 to 4 arrive and are all stored in the cold queue.
- request A 1 B 1 for establishment of a SAQ for storage of packets destined for the route A 1 B 1 arrives. There are no existing SAQs and so a SAQ for A 1 B 1 is established.
- An activating link (R/L A 1 B 1 ) to the A 1 B 1 SAQ is placed in the cold queue. Packets 10 and 11 that arrive subsequently are placed in the SAQ A 1 B 1 . Packets 12 to 14 arrive subsequently and are stored in the cold queue.
- request A 1 B 2 arrives. There are no existing SAQs with which the request A 1 B 2 clashes i.e. no SAQ A 1 B 2 is already formed so a SAQ A 1 B 2 is formed. An activating link (R/L A 1 B 2 ) to the SAQ A 1 B 2 is placed in the cold queue. Subsequently, packets 20 to 23 arrive and these are placed in SAQs A 1 B 1 and A 1 B 2 accordingly.
- request A 1 arrives. This is a less specific request than some existing SAQs (i.e. A 1 B 1 and A 1 B 2 ).
- some existing SAQs i.e. A 1 B 1 and A 1 B 2 .
- one of the CAM lines corresponding to the SAQs A 1 B 1 and A 1 B 2 is overwritten, i.e. replaced with a new CAM line A 1 .
- the shortest existing SAQ is selected for overwriting i.e. the SAQ corresponding to the CAM line entry having the smallest number of stages. If, as in this case, there is no shortest SAQ then any (either in this case) may be picked from the existing shorter SAQs. Since the new SAQ is shorter than the existing SAQ being overwritten, data packets already assigned to the existing SAQ do not need to be moved. This is because had they have arrived when only the new shorter SAQ existed, they would have been placed in that shorter SAQ anyway.
- the CAM entry A 1 B 2 is overwritten without the requirement to move any packets in the SAQ A 1 B 2 .
- a link (R/L A 1 ) to the A 1 SAQ is placed in the cold queue. If it still exists in the cold queue, the existing link A 1 B 2 must now be ignored. This is achieved by providing links with sequence numbers so that only the highest matching link can activate a SAQ.
- Packets 30 to 34 subsequently arrive and these are placed in SAQs A 1 B 1 and A 1 accordingly, i.e. packets 30 and 31 are placed in SAQ A 1 B 1 and packets 32 to 34 are placed in SAQ A 1 .
- a request A 1 B 1 C 1 arrives. This is a more specific request than all the existing CAM lines and SAQs. Again, one of the existing CAM lines corresponding to an SAQ is overwritten. The shortest SAQ is selected. In this case this is SAQ A 1 . To avoid out of order effects the existing A 1 SAQ is linked to the cold queue via a link 92 . In effect, the data packets stored in the SAQ A 1 are moved to the end of the cold queue. If this were not done, a new data packet A 1 B 2 C 1 , e.g.
- packet number 42 may be selected for transmission from the port by the scheduler before the packets A 1 B 2 C 1 (packets 22 and 32 ) stored in the SAQ A 1 .
- SAQ A 1 contains all data packets previously assigned to the SAQ A 1 B 2 .
- An activating link 94 to the SAQ A 1 B 1 C 1 is placed in the existing SAQ which is the next longest SAQ. In this case this is SAQ A 1 B 1 . Packets 40 and 41 subsequently arrive and these are placed in the corresponding SAQs (A 1 B 1 C 1 and A 1 B 1 respectively).
Abstract
Description
Claims (35)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US10/590,209 US8174978B2 (en) | 2004-03-05 | 2005-03-04 | Method for congestion management of a network, a signalling protocol, a switch, an end station and a network |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US55005804P | 2004-03-05 | 2004-03-05 | |
US62299504P | 2004-10-29 | 2004-10-29 | |
US10/590,209 US8174978B2 (en) | 2004-03-05 | 2005-03-04 | Method for congestion management of a network, a signalling protocol, a switch, an end station and a network |
PCT/GB2005/000836 WO2005086435A1 (en) | 2004-03-05 | 2005-03-04 | A method for congestion management of a network, a signalling protocol, a switch, an end station and a network |
Publications (2)
Publication Number | Publication Date |
---|---|
US20080253289A1 US20080253289A1 (en) | 2008-10-16 |
US8174978B2 true US8174978B2 (en) | 2012-05-08 |
Family
ID=34922726
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US10/590,209 Active 2027-06-23 US8174978B2 (en) | 2004-03-05 | 2005-03-04 | Method for congestion management of a network, a signalling protocol, a switch, an end station and a network |
Country Status (5)
Country | Link |
---|---|
US (1) | US8174978B2 (en) |
EP (1) | EP1728366B1 (en) |
JP (1) | JP2007526706A (en) |
DE (1) | DE602005003652D1 (en) |
WO (1) | WO2005086435A1 (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9197563B2 (en) | 2013-03-15 | 2015-11-24 | International Business Machines Corporation | Bypassing congestion points in a converged enhanced ethernet fabric |
US9219691B2 (en) | 2013-03-15 | 2015-12-22 | International Business Machines Corporation | Source-driven switch probing with feedback request |
US9401857B2 (en) | 2013-03-15 | 2016-07-26 | International Business Machines Corporation | Coherent load monitoring of physical and virtual networks with synchronous status acquisition |
US9497125B2 (en) | 2013-07-28 | 2016-11-15 | Mellanox Technologies Ltd. | Congestion control enforcement in a virtualized environment |
US9807024B2 (en) | 2015-06-04 | 2017-10-31 | Mellanox Technologies, Ltd. | Management of data transmission limits for congestion control |
US9954781B2 (en) | 2013-03-15 | 2018-04-24 | International Business Machines Corporation | Adaptive setting of the quantized congestion notification equilibrium setpoint in converged enhanced Ethernet networks |
US10009277B2 (en) | 2015-08-04 | 2018-06-26 | Mellanox Technologies Tlv Ltd. | Backward congestion notification in layer-3 networks |
US10237376B2 (en) | 2015-09-29 | 2019-03-19 | Mellanox Technologies, Ltd. | Hardware-based congestion control for TCP traffic |
Families Citing this family (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8509077B2 (en) * | 2006-04-05 | 2013-08-13 | Xyratex Technology Limited | Method for congestion management of a network, a switch, and a network |
US8154996B2 (en) | 2008-09-11 | 2012-04-10 | Juniper Networks, Inc. | Methods and apparatus for flow control associated with multi-staged queues |
US8325749B2 (en) | 2008-12-24 | 2012-12-04 | Juniper Networks, Inc. | Methods and apparatus for transmission of groups of cells via a switch fabric |
US8213308B2 (en) | 2008-09-11 | 2012-07-03 | Juniper Networks, Inc. | Methods and apparatus for defining a flow control signal related to a transmit queue |
US8254255B2 (en) * | 2008-12-29 | 2012-08-28 | Juniper Networks, Inc. | Flow-control in a switch fabric |
US9264321B2 (en) | 2009-12-23 | 2016-02-16 | Juniper Networks, Inc. | Methods and apparatus for tracking data flow based on flow state values |
US9065773B2 (en) | 2010-06-22 | 2015-06-23 | Juniper Networks, Inc. | Methods and apparatus for virtual channel flow control associated with a switch fabric |
US9660940B2 (en) | 2010-12-01 | 2017-05-23 | Juniper Networks, Inc. | Methods and apparatus for flow control associated with a switch fabric |
US9032089B2 (en) | 2011-03-09 | 2015-05-12 | Juniper Networks, Inc. | Methods and apparatus for path selection within a network based on flow duration |
US8811183B1 (en) | 2011-10-04 | 2014-08-19 | Juniper Networks, Inc. | Methods and apparatus for multi-path flow control within a multi-stage switch fabric |
US20130111084A1 (en) * | 2011-11-01 | 2013-05-02 | Qualcomm Incorporated | Methods and devices for single link aggregated buses |
JP6007595B2 (en) * | 2012-05-28 | 2016-10-12 | 富士通株式会社 | Transmission method, apparatus and program |
US8984206B2 (en) | 2012-10-31 | 2015-03-17 | International Business Machines Corporation | Weightage-based scheduling for hierarchical switching fabrics |
US8902899B2 (en) | 2013-02-08 | 2014-12-02 | International Business Machines Corporation | Input buffered switching device including bypass logic |
JP6281327B2 (en) * | 2014-03-06 | 2018-02-21 | 富士通株式会社 | Information processing system, information processing apparatus, switch apparatus, and information processing system control method |
WO2015142241A1 (en) * | 2014-03-20 | 2015-09-24 | Telefonaktiebolaget L M Ericsson (Publ) | Tunnel congestion volume policing |
US9467396B2 (en) | 2014-04-11 | 2016-10-11 | International Business Machines Corporation | Simultaneous transfers from a single input link to multiple output links with a timesliced crossbar |
CN105991471B (en) * | 2015-02-16 | 2019-08-09 | 新华三技术有限公司 | The flow control method and flow control apparatus and the network equipment of the network equipment |
CN113728596A (en) | 2019-05-23 | 2021-11-30 | 慧与发展有限责任合伙企业 | System and method for facilitating efficient management of idempotent operations in a Network Interface Controller (NIC) |
CN112565087A (en) * | 2020-11-23 | 2021-03-26 | 盛科网络(苏州)有限公司 | PFC (Power factor correction) back-pressure message and processing method thereof |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0717532A1 (en) | 1994-12-13 | 1996-06-19 | International Business Machines Corporation | Dynamic fair queuing to support best effort traffic in an ATM network |
US20030185154A1 (en) * | 2002-03-29 | 2003-10-02 | Nishan Systems, Inc. | Network congestion management systems and methods |
US20030193959A1 (en) * | 2002-01-20 | 2003-10-16 | General Instrument Corporation | Method and apparatus for priority-based load balancing for use in an extended local area network |
US6657963B1 (en) * | 1999-07-30 | 2003-12-02 | Alcatel Canada Inc. | Method and apparatus for controlling data congestion in a frame relay/ATM internetworking system |
US20050088969A1 (en) * | 2001-12-19 | 2005-04-28 | Scott Carlsen | Port congestion notification in a switch |
US20050147032A1 (en) * | 2003-12-22 | 2005-07-07 | Lyon Norman A. | Apportionment of traffic management functions between devices in packet-based communication networks |
WO2006032978A2 (en) * | 2004-09-22 | 2006-03-30 | Freesystems Pte., Ltd. | A method and apparatus for ensuring high quality audio playback in a wireless or wired digital audio communication system |
US7286552B1 (en) * | 2002-07-26 | 2007-10-23 | Extreme Networks | Method and apparatus for providing quality of service across a switched backplane for multicast packets |
US20100118703A1 (en) * | 2004-06-04 | 2010-05-13 | David Mayhew | System and method to identify and communicate congested flows in a network fabric |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0712220A1 (en) * | 1994-11-08 | 1996-05-15 | International Business Machines Corporation | Hop-by-hop flow control in an ATM network |
-
2005
- 2005-03-04 WO PCT/GB2005/000836 patent/WO2005086435A1/en active Application Filing
- 2005-03-04 JP JP2007501353A patent/JP2007526706A/en active Pending
- 2005-03-04 EP EP05717906A patent/EP1728366B1/en not_active Expired - Fee Related
- 2005-03-04 DE DE602005003652T patent/DE602005003652D1/en active Active
- 2005-03-04 US US10/590,209 patent/US8174978B2/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0717532A1 (en) | 1994-12-13 | 1996-06-19 | International Business Machines Corporation | Dynamic fair queuing to support best effort traffic in an ATM network |
US6657963B1 (en) * | 1999-07-30 | 2003-12-02 | Alcatel Canada Inc. | Method and apparatus for controlling data congestion in a frame relay/ATM internetworking system |
US20050088969A1 (en) * | 2001-12-19 | 2005-04-28 | Scott Carlsen | Port congestion notification in a switch |
US20030193959A1 (en) * | 2002-01-20 | 2003-10-16 | General Instrument Corporation | Method and apparatus for priority-based load balancing for use in an extended local area network |
US20030185154A1 (en) * | 2002-03-29 | 2003-10-02 | Nishan Systems, Inc. | Network congestion management systems and methods |
US7286552B1 (en) * | 2002-07-26 | 2007-10-23 | Extreme Networks | Method and apparatus for providing quality of service across a switched backplane for multicast packets |
US20050147032A1 (en) * | 2003-12-22 | 2005-07-07 | Lyon Norman A. | Apportionment of traffic management functions between devices in packet-based communication networks |
US20100118703A1 (en) * | 2004-06-04 | 2010-05-13 | David Mayhew | System and method to identify and communicate congested flows in a network fabric |
WO2006032978A2 (en) * | 2004-09-22 | 2006-03-30 | Freesystems Pte., Ltd. | A method and apparatus for ensuring high quality audio playback in a wireless or wired digital audio communication system |
Non-Patent Citations (3)
Title |
---|
International Preliminary Report on Patentability for application PCT/GB2005/000836. |
Schroeder et al., "Autonet: A High-Speed, Self-Configuring Local Area Network Using Point-to-Point Links," IEEE Journal on Selected Areas in Communications, IEEE Inc., New York, U.S., vol. 9, No. 8, Oct. 1, 1991, pp. 1318-1335. |
Written Opinion of the International Search Authority in application PCT/GB2005/000836. |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9197563B2 (en) | 2013-03-15 | 2015-11-24 | International Business Machines Corporation | Bypassing congestion points in a converged enhanced ethernet fabric |
US9219691B2 (en) | 2013-03-15 | 2015-12-22 | International Business Machines Corporation | Source-driven switch probing with feedback request |
US9219689B2 (en) | 2013-03-15 | 2015-12-22 | International Business Machines Corporation | Source-driven switch probing with feedback request |
US9253096B2 (en) | 2013-03-15 | 2016-02-02 | International Business Machines Corporation | Bypassing congestion points in a converged enhanced ethernet fabric |
US9401857B2 (en) | 2013-03-15 | 2016-07-26 | International Business Machines Corporation | Coherent load monitoring of physical and virtual networks with synchronous status acquisition |
US9954781B2 (en) | 2013-03-15 | 2018-04-24 | International Business Machines Corporation | Adaptive setting of the quantized congestion notification equilibrium setpoint in converged enhanced Ethernet networks |
US9998377B2 (en) | 2013-03-15 | 2018-06-12 | International Business Machines Corporation | Adaptive setting of the quantized congestion notification equilibrium setpoint in converged enhanced ethernet networks |
US9497125B2 (en) | 2013-07-28 | 2016-11-15 | Mellanox Technologies Ltd. | Congestion control enforcement in a virtualized environment |
US9807024B2 (en) | 2015-06-04 | 2017-10-31 | Mellanox Technologies, Ltd. | Management of data transmission limits for congestion control |
US10009277B2 (en) | 2015-08-04 | 2018-06-26 | Mellanox Technologies Tlv Ltd. | Backward congestion notification in layer-3 networks |
US10237376B2 (en) | 2015-09-29 | 2019-03-19 | Mellanox Technologies, Ltd. | Hardware-based congestion control for TCP traffic |
Also Published As
Publication number | Publication date |
---|---|
JP2007526706A (en) | 2007-09-13 |
WO2005086435A1 (en) | 2005-09-15 |
US20080253289A1 (en) | 2008-10-16 |
EP1728366A1 (en) | 2006-12-06 |
EP1728366B1 (en) | 2007-12-05 |
DE602005003652D1 (en) | 2008-01-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8174978B2 (en) | Method for congestion management of a network, a signalling protocol, a switch, an end station and a network | |
US11818037B2 (en) | Switch device for facilitating switching in data-driven intelligent network | |
US7957293B2 (en) | System and method to identify and communicate congested flows in a network fabric | |
US7324460B2 (en) | Event-driven flow control for a very high-speed switching node | |
US20020141427A1 (en) | Method and apparatus for a traffic optimizing multi-stage switch fabric network | |
US8509077B2 (en) | Method for congestion management of a network, a switch, and a network | |
JP2001285364A (en) | Switching apparatus and method for same | |
US5699520A (en) | Flow control apparatus and method for a computer interconnect using adaptive credits and flow control tags | |
JP2008166888A (en) | Priority band control method in switch | |
US6389017B1 (en) | Resource scheduling algorithm in packet switched networks with multiple alternate links | |
Gómez et al. | VOQ/sub SW: a methodology to reduce HOL blocking in InfiniBand networks | |
US11805080B2 (en) | System and method for data loss and data latency management in a network-on-chip with buffered switches | |
US20240056385A1 (en) | Switch device for facilitating switching in data-driven intelligent network | |
Pascoe et al. | Configuration Discovery and Mapping of a Home Network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: VIRTENSYS LIMITED, UNITED KINGDOM Free format text: LICENSE;ASSIGNOR:XYRATEX TECHNOLOGY LIMITED;REEL/FRAME:018951/0854 Effective date: 20070108 |
|
AS | Assignment |
Owner name: XYRATEX TECHNOLOGY LIMITED, UNITED KINGDOM Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:NAVEN, FINBAR;JOHNSON, IAN DAVID;DUATO, JOSE;AND OTHERS;REEL/FRAME:021162/0042;SIGNING DATES FROM 20070617 TO 20070621 Owner name: XYRATEX TECHNOLOGY LIMITED, UNITED KINGDOM Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:NAVEN, FINBAR;JOHNSON, IAN DAVID;DUATO, JOSE;AND OTHERS;SIGNING DATES FROM 20070617 TO 20070621;REEL/FRAME:021162/0042 |
|
AS | Assignment |
Owner name: KREOS CAPITAL III (LUXEMBOURG) S.A.R.L., LUXEMBOUR Free format text: SECURITY AGREEMENT;ASSIGNOR:VIRTENSYS LIMITED;REEL/FRAME:025126/0776 Effective date: 20100922 |
|
FEPP | Fee payment procedure |
Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
AS | Assignment |
Owner name: MICRON TECHNOLOGY, INC, IDAHO Free format text: RELEASE BY SECURED PARTY;ASSIGNOR:KREOS CAPITAL III (LUXEMBOURG) S.A.R.L;REEL/FRAME:028233/0608 Effective date: 20120131 |
|
FPAY | Fee payment |
Year of fee payment: 4 |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 8TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1552); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 8 |
|
FEPP | Fee payment procedure |
Free format text: MAINTENANCE FEE REMINDER MAILED (ORIGINAL EVENT CODE: REM.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |